Update app.py

app.py

@@ -28,9 +28,12 @@ logger = logging.getLogger(__name__)
 
 API_KEY = "sk-your-secret-key"
 
-# timeout لمنع التعليق الأبدي
+# timeout عام
 REQUEST_TIMEOUT = 45
 
+# timeout للـ stream نفسه
+STREAM_CHUNK_TIMEOUT = 20
+
 # retry خفيف
 MAX_RETRIES = 2
 
@@ -41,7 +44,7 @@ MAX_RETRIES = 2
 
 app = FastAPI(
     title="Universal AI Gateway",
-    version="4.2.0"
+    version="4.3.0"
 )
 
 
@@ -114,7 +117,7 @@ async def root():
     return {
         "status": "online",
         "service": "Universal AI Gateway",
-        "version": "4.2.0"
+        "version": "4.3.0"
     }
 
 
@@ -128,7 +131,7 @@ async def get_models():
     models_data = []
 
     # =================================================
-    # MODELS THAT WORK WELL
+    # CURATED MODELS
     # =================================================
 
     fallback_models = [
@@ -141,18 +144,36 @@ async def get_models():
 
         # Claude
         "claude-3-haiku",
+        "claude-3-sonnet",
 
         # Llama
         "llama-3.1-70b",
+        "llama-3.1-8b",
+
+        # Qwen
+        "qwen-2-72b",
+        "qwen-2.5-72b",
+
+        # DeepSeek
+        "deepseek-chat",
+        "deepseek-v3",
+        "deepseek-r1-distill-qwen-14b",
+
+        # Kimi
+        "kimi-k2",
 
         # Mixtral
         "mixtral-8x7b",
 
-        # Deepseek
-        "deepseek-chat",
+        # Aria
+        "aria",
+
+        # Sonar
+        "sonar-reasoning",
+        "sonar-reasoning-pro",
 
-        # Gemini
-        "gemini-pro",
+        # Nemotron
+        "nemotron-70b",
 
         # =================================================
         # COHERE FAMILY
@@ -162,11 +183,17 @@ async def get_models():
         "command-r-plus",
         "command-r7b",
         "command",
+        "command-light",
+        "command-light-nightly",
         "command-nightly",
 
-        # Additional Cohere-style names
+        # Cohere advanced
         "cohere-command-r",
         "cohere-command-r-plus",
+
+        # Additional aliases
+        "c4ai-command-r-plus",
+        "c4ai-command-r-v01",
     ]
 
     added_models = set()
@@ -177,7 +204,7 @@ async def get_models():
 
             all_models = list(g4f.models._all_models)
 
-            for model in all_models[:100]:
+            for model in all_models[:120]:
 
                 model_name = str(model)
 
@@ -196,7 +223,10 @@ async def get_models():
 
         logger.error(f"Models error: {e}")
 
-    # fallback models
+    # =================================================
+    # ADD CURATED MODELS
+    # =================================================
+
     for model in fallback_models:
 
         if model not in added_models:
@@ -212,7 +242,8 @@ async def get_models():
 
     return {
         "object": "list",
-        "data": models_data
+        "data": models_data,
+        "total": len(models_data)
     }
 
 
@@ -238,7 +269,6 @@ async def safe_completion(
 
             client = Client()
 
-            # timeout لمنع التعليق الأبدي
             response = await asyncio.wait_for(
                 asyncio.to_thread(
                     client.chat.completions.create,
@@ -276,6 +306,15 @@ async def safe_completion(
     raise Exception(last_error)
 
 
+# =====================================================
+# STREAM WATCHDOG
+# =====================================================
+
+async def next_chunk(iterator):
+
+    return await asyncio.to_thread(next, iterator)
+
+
 # =====================================================
 # CHAT COMPLETIONS
 # =====================================================
@@ -320,12 +359,52 @@ async def chat_completions(
 
                 has_content = False
 
-                for chunk in response:
+                while True:
+
+                    try:
+
+                        # watchdog للـ stream
+                        chunk = await asyncio.wait_for(
+                            next_chunk(response),
+                            timeout=STREAM_CHUNK_TIMEOUT
+                        )
+
+                    except StopIteration:
+                        break
+
+                    except asyncio.TimeoutError:
+
+                        logger.warning(
+                            f"Stream stalled | model={body.model}"
+                        )
+
+                        error_payload = {
+                            "error": {
+                                "message": "Stream timeout",
+                                "type": "stream_timeout"
+                            }
+                        }
+
+                        yield (
+                            f"data: "
+                            f"{json.dumps(error_payload)}\n\n"
+                        )
+
+                        break
 
                     try:
 
                         content = ""
 
+                        # حماية من ImageResponse
+                        if hasattr(chunk, "images"):
+
+                            logger.warning(
+                                f"Image model detected | model={body.model}"
+                            )
+
+                            continue
+
                         if (
                             hasattr(chunk, "choices")
                             and chunk.choices
@@ -367,7 +446,7 @@ async def chat_completions(
                             f"Chunk error: {chunk_error}"
                         )
 
-                # provider فتح stream بدون محتوى
+                # stream فارغ
                 if not has_content:
 
                     error_payload = {